Pathfinder AI

List of Visa
Landing Page
Profile analysis

INSPIRATION :

The idea for Pathfinder AI was born from a personal frustration many of us face: the overwhelming complexity of international relocation. With over 280 million international migrants worldwide, the process of understanding visa eligibility, finding jobs with sponsorship, and locating suitable housing remains fragmented across dozens of websites, government portals, and forums. We asked ourselves: What if AI could be your personal immigration consultant? One that analyzes your unique profile, understands visa requirements across multiple countries, and delivers a personalized roadmap—all in under 60 seconds.

WHAT IT DOES

Pathfinder AI is an intelligent relocation assistant that:

Evaluates Visa Eligibility - Using a hybrid AI model that combines rule-based logic with Google's Gemini AI to analyze 50+ visa types across 8 countries
Calculates Success Probability - Provides a data-driven assessment of your relocation chances
Recommends Job Opportunities - Curates positions with visa sponsorship that match your profile
Suggests Housing Options - Finds accommodations in your target destination
Provides Cost of Living Analysis - Breaks down monthly expenses so you can plan financially

HOW WE BUILT IT

We architected Pathfinder AI as a modern full-stack application: Backend Architecture:

Built a REST API with FastAPI for high-performance async operations
Implemented a two-stage hybrid AI model:
- Stage 1: Deterministic rule engine evaluating profiles against visa requirements
- Stage 2: Gemini AI for deep analysis, personalized recommendations, and natural language insights
Structured data using Pydantic models for type safety and validation Frontend Experience:
Created an immersive UI with React 18 and TypeScript
Designed a custom dark theme with glassmorphism effects using Material-UI
Implemented fluid animations with Framer Motion - floating orbs, page transitions, and micro-interactions
Built data visualizations using react-circular-progressbar and custom components AI Integration:
Engineered prompts for Google Gemini API to return structured JSON responses
Implemented fallback mechanisms for API resilience

CHALLENGES WE RAN INTO

Prompt Engineering for Structured Output - Getting Gemini to consistently return valid JSON required careful prompt design and response parsing with fallback handling
Visa Rule Complexity - Each country has unique visa categories with different requirements. We had to normalize this data into a queryable format while preserving accuracy
Type Safety Across Stack - Ensuring TypeScript interfaces matched Pydantic models required careful coordination between frontend and backend

ACCOMPLISHMENTS THAT WE ARE PROUD OF

Built a production-ready full-stack application in a hackathon timeframe
Achieved sub-2-second AI analysis response times
Created a UI that feels like a premium SaaS product, not a hackathon prototype
Implemented a hybrid AI approach that combines deterministic rules with generative AI

WHAT WE LEARNED

The power of combining rule-based systems with LLMs for reliable, explainable AI
Advanced prompt engineering techniques for structured data extraction
Building performant animations that enhance rather than hinder UX
The importance of graceful degradation when external APIs are unavailable

WHAT'S NEXT